A DARPA Perspective on AI · PDF fileDARPA Autonomous Vehicle Grand Challenge. 140 miles of...
-
Upload
nguyenkien -
Category
Documents
-
view
233 -
download
0
Transcript of A DARPA Perspective on AI · PDF fileDARPA Autonomous Vehicle Grand Challenge. 140 miles of...
Artificial intelligence
The first wave is still advancing and solving hard problems
The second wave is amazingly effective, but it has fundamental limitations
New research is shaping the third wave
Three waves of AI technology (so far)
Statistical learning Contextual adaptation Handcrafted knowledge
Intelligence is an ability to process information
perceive rich, complex and subtle information
learn
within an environment
abstract to create new meanings
reason
to plan and to decide
Artificial intelligence is a programmed ability to process information
perceiving
abstracting reasoning
learning
Intelligence scale
First-wave AI technologies
Planning tools
Expert systems Cybersecurity
Command Post of the Future
Handcrafted knowledge
Perceiving
Abstracting Reasoning
Learning
Engineers create sets of rules to represent knowledge in well defined domains AI systems reason over narrowly defined problems No learning capability and poor handling of uncertainty
First wave stumbles on natural data
DARPA Autonomous Vehicle Grand Challenge 140 miles of dirt tracks in California and Nevada
2004 # completed: 0
2005 # completed: 5
The problem in 2004
Vehicles were able to follow the GPS waypoints very accurately
but either missed or hallucinated obstacles ahead
The difference in 2005
A probabilistic algorithm interpreted the flood of incoming sensor data to
learn the terrain and map out an optimal driving surface
Second-wave AI technologies
Image recognition
Text analysis
Apple Siri
AlphaGo
Statistical learning
Engineers create sets of rules to represent knowledge in well defined domains AI systems reason over narrowly defined problems No learning capability and poor handling of uncertainty
Engineers create statistical models for specific problem domains and train them on big data AI systems have nuanced classification and prediction capabilities No contextual capability and minimal reasoning ability
Handcrafted knowledge
Perceiving
Abstracting Reasoning
Learning Perceiving
Abstracting Reasoning
Learning
Farfade, Saberian, and Li 2015
Patents 2004-2013
Neural Nets
Statistical Models
Ensemble Methods
Decision Trees
Deep Learning
SVMs
AOGs
Bayesian Belief Nets
Markov Models
HBNs
MLNs
SRL CRFs
Random Forests
Graphical Models
Key enablers of second-wave AI
130 1,227
7,910
2005 2010
2015
4,400 44,000 exabytes
2013
2020
Big data
Power-efficient processing
Many applications and big markets
Better machine-learning algorithms
1
10
100
1000
2000 2005 2010 2015 2020
CPU
GPU
KW for 1B- parameter
deep learning network
Operate highly autonomous platforms
Today’s artificial intelligence is powerful…
Search the deep web
Manage cybersecurity in real time
Overcome spectrum scarcity
Andrej Karpathy, Li Fei-Fei
…but limited
Construction worker in orange safety vest is working on road
A young boy is holding a baseball bat
Future third-wave AI technologies
Statistical learning Contextual adaptation
Engineers create systems that construct explanatory models for classes of real-world phenomena AI systems learn and reason as they encounter new tasks and situations Natural communication among machines and people
Engineers create sets of rules to represent knowledge in well defined domains AI systems reason over narrowly defined problems No learning capability and poor handling of uncertainty
Engineers create statistical models for specific problem domains and train them on big data AI systems have nuanced classification and prediction capabilities No contextual capability and minimal reasoning ability
Handcrafted knowledge
Perceiving
Abstracting Reasoning
Learning Perceiving
Abstracting Reasoning
Learning Perceiving
Abstracting Reasoning
Learning
Develop explainable AI
New learning process
Training data Explainable model
Explanation interface
This is a cat: • It has fur, whiskers,
and claws. • It has this feature:
• I understand why/why not
• I know when it will succeed/fail
Early result: An automatically generated web of influences from 1000 research papers
Uncover causal relationships among 1,000,000s of scientific observations
Explainable AI
Continuous learning
Embedded machine learning
Automatic whole-system causal models
Human-machine symbiosis
Some third-wave AI technologies
R&D investments in foundational technology and applications
Domain: Natural language processing
GALE: Over 35 systems deployed worldwide
Apple Siri Speech Understanding Research (1971)
IBM Watson
Robust Automatic Transcription of Speech (2010)
Dragon (Nuance NaturallySpeaking)
Low Resource Languages for Emergent Incidents (2015)
Big Mechanism (2015)
Broad Operational Language Translation (2012)
Communicating with Computers (2015)
Deep Exploration and Filtering of Text (2012)
Memex (2013)
Facebook AI (LeCun)
Baidu Research (Ng)
Artificial Neural Nets (1991)
XDATA (2011)
Global Autonomous Language Exploitation (2005)
Machine Reading (2005)
Personalized Assistant that Learns (2005)
Deep Learning (2010)
USA CPoF, USSTRATCOM, NMCI
DTRA, NASIC
AF Rivet Joint, SOCOM PUMA UAV, IC, JSOC Traveler, JIATF-S, Navy EP3
Military Foreign language Translation System POR, 16 DoD organizations, FBI
DNI, CIA, Treasury, FinCEN, FBI, DEA
Law Enforcement, ATF, FBI, NSA
DTRA terrorism indicators
DARPA programs
Direct commercial impact
DoD transition